Message dissemination in vehicular ad hoc networks

Abstract

A vehicular network is a challenging environment since it combines a xed infrastructure(roadside units, e.g., proxies), and ad hoc communications among vehicles. Vehicularnetworks have the diverse range of applications that varies from safety applications to comfortapplications. As the dissemination requirements (reliability, delay, coverage, etc.) maybe dierent from one application to another, the dissemination mechanism could be implementedat the application layer where the interaction with the routing layer is limited. Thisthesis addresses several aspects of vehicular communications related to data dissemination.This thesis concerns with centrality metrics in ad hoc networks. We present a centralitymetric that characterize the network topology using only limited, local connectivityinformation one or two hop information. We quantify a nodes' centrality based on thenumber of neighbors a node belongs to and the ranking this node has in each one. Thatway nodes that belong to many clusters and participate in many communication paths isranked high. No nodes remain unranked while for large networks the method makes similar to PageRank rankings, even though it is a localized metric whereas PageRank requirescumbersome computations and knowledge of the whole networks topology.In the present work we also create novel clustering methods that work ecientlyboth in highway and urban environments. The methods combine dierent information suchas current and future possible position of vehicles, relative mobility, vehicles' direction andheight in order to create stable clusters. Historical data of drivers are also used in order togroup those that share common habits and thus mobility patterns. Social aspects of driving are combined with mathematically measurable parameters for cluster stability, with the useof DSRC communication capabilities.This thesis addresses the issue of delay constraints in packet scheduling routing in both static and mobile ad hoc networks. We present novel methods for delay ecient routingof packets based on the well known Back-pressure routing protocol. Large delays that appearwhen network load is low or medium, are addressed with the combination of packets' travelhistory or with the segmentation of the network in clusters. Both methods manage toperform eectively in terms of mean delay, packet throughput and energy consumption.A Push-based data dissemination mechanism that use xed RSUs is also presentedthat addresses the issue of skewed access pattern. In real life applications, some items aremore frequently accessed by clients than some others and the correct indexing of these datais very important. The method exploits the dierent access probabilities of data items, inorder to improve the mean tuning time, while retaining the same access time low. Thepresence of air indexes would give an answer about the time of broadcast and is appropriatefor helping the driver choose the correct driving behavior in order to retrieve this neededinformation.We also present a pull-based congestion avoidance method for VANETs. We proposea real time system based on DSRC communication capabilities in order to reroute vehiclesto the most ecological route, avoiding congested roads and balancing overall travel timeand C02 emissions. The proposed method uses real time information broadcasted frompassing vehicles to RSUs scattered along the road in order to assign weights to dierentroad segments according to time, distance and CO2 emissions. The method can be used fortrac management situations that arise after an accident in a city.